A journal I am submitting to requires camera-ready figures. What does this term mean nowadays? Does it imply high resolution? My figures are in png format and have 96 dpi. Enlarging it can produce some jagged edges. Should I update them to eps or pdf format?
There isn't any real, single standard as to what "camera-ready figures" means. Different journals/publishers may use different publishing processes and workflows, which can result in different requirements for file formats, resolution etc. (Or even color spaces.) However, 96 dpi is indeed quite low. 300 dpi is a more reasonable minimum guideline for raster images - but obviously vector graphics should be preferred whenever possible.
In lieu of a standard, we can look at typical recommendations. A review by J.H. Lee in Science Editing titled Handling digital images for publication states that
The most commonly recommended resolution for printing on paper depends on the nature of the images: 1) 300 dpi for color pictures, 2) 300 to 600 dpi for black and white pictures, 3) 600 to 900 dpi for combination art (photo and text), and 4) 900 to 1,200 dpi for line art.
Note that this is the printed dpi, which isn't necessarily the same dpi as seen in the submitted manuscript. The same author also suggests a universal guide to avoid issues with publishers rescaling figures etc.:
It is the opinion of this author that a universal recommendation could help authors prepare their images. The standard figure size of most academic journals is about 86 mm (single column). The standard pixels per inch for line art is 900 to 1,200 ppi. Therefore, an image file of 900 ppi and 4 inches is of sufficient quality for most publications; this means 3,600 pixels in a horizontal line. It is recommended that authors use this number as a universal guide.
Here is a rantswer: In my experience (limited to my field) it means "We're a cheap journal that doesn't care about graphics quality, so we let the authors do all the work even if the figures will be subpar." I have never had one of these journals complain about the quality of my figures.
On the other hand, when checking galley proofs, I have complained a few times that my Tikz figures ended up having the wrong font (because the production team changed the manuscript font but copypasted the figures rather than recompiling them). And, guess what, nothing happened.